YouTube Launches Beta Test of AI-Powered Playables Builder

YouTube has begun a closed beta test of a new tool called Playables Builder, which allows users to create simple video games using artificial intelligence, with no need to write code. The tool, presented as a web app prototype built on Gemini 3, operates on a "enter a prompt, get a game" principle.

Users can generate a game by uploading a short video of desired gameplay, an image, or simply by describing their idea in text. At this stage, access to the builder is limited to select creators, and the games they've already created are available to launch directly on YouTube, but are geographically restricted: they can be tried by users in the US, Canada, the UK, and Australia.

At the core of Playables Builder is Google's new AI model, Gemini 3, which the company calls its "most intelligent" to date. A key capability of the model used in this project is so-called "vibe coding" — the ability to transform informal descriptions, sketches, or multimedia prompts into working code. The first generated games are extremely simple arcade games, platformers, and puzzles.

Experts are skeptical about the depth and engagement of this approach to game development, with the main criticism being that creating a truly compelling game requires iteration, fine-tuning, and expertise that AI at this stage cannot replace.

The launch of the game builder is part of Google's broader strategy to integrate AI into its services. YouTube previously tested small games (Playables) on the platform and added multiplayer to them last year. Google has not yet announced a timeline for a public release of Playables Builder. The success or failure of this initiative could show how ready content creators and their audiences are to embrace games generated by a neural network from a text description.